Train LORA model with ComfyUI Locally & FREE & no code For AI Image Generation

Опубликовано: 20 Февраль 2026
на канале: Ao Ao University AI

5,187

167

comfyui
https://github.com/comfyanonymous/Com...

comfyui manager
https://github.com/Comfy-Org/ComfyUI-...

comfyui FLUX trainer
https://github.com/kijai/ComfyUI-Flux...

FLUX dev model
https://huggingface.co/lllyasviel/flu...

Text encoders (Clips)
https://huggingface.co/comfyanonymous...
https://huggingface.co/comfyanonymous...

VAE
https://huggingface.co/StableDiffusio...

The trained LORA checkpoints are by default in the folder "flux_lora_output_path". Use the one without "-step0xxxxx". For example: my checkpoint is "flux_lora_aoao2_rank16_bf16.safetensors", which is the newest updated one in the folder.
It took 10 min to train 4 photos with 100 steps per loop.
For 12 photos and 750 steps per loop took 2hrs 45min. I am using rtx4090 + amd 9950x +64gb